Report standard errors in README.md

This commit is contained in:
Felix Bauckholt 2019-03-09 14:41:18 +01:00 committed by Jeff Wu
parent 96da95dbba
commit 8fd230da1c
3 changed files with 41 additions and 14 deletions

View file

@ -69,9 +69,9 @@ time cargo run --release -- --write-results-table
On the first 20000 seeds, we have these average scores and win rates: On the first 20000 seeds, we have these average scores and win rates:
| | 2p | 3p | 4p | 5p | | | 2p (stderr) | 3p (stderr) | 4p (stderr) | 5p (stderr) |
|---------|---------|---------|---------|---------| |---------|------------------|------------------|------------------|------------------|
| cheat | 24.8594 | 24.9785 | 24.9720 | 24.9557 | | cheat | 24.8594 (0.0036) | 24.9785 (0.0012) | 24.9720 (0.0014) | 24.9557 (0.0018) |
| | 90.59 % | 98.17 % | 97.76 % | 96.42 % | | | 90.59 % (0.21 %) | 98.17 % (0.09 %) | 97.76 % (0.10 %) | 96.42 % (0.13 %) |
| info | 22.3249 | 24.7278 | 24.8919 | 24.8961 | | info | 22.3249 (0.0128) | 24.7278 (0.0046) | 24.8919 (0.0029) | 24.8961 (0.0027) |
| | 09.81 % | 80.54 % | 91.67 % | 91.90 % | | | 09.81 % (0.21 %) | 80.54 % (0.28 %) | 91.67 % (0.20 %) | 91.90 % (0.19 %) |

View file

@ -166,12 +166,13 @@ fn get_results_table() -> String {
let n_threads = 8; let n_threads = 8;
let intro = format!("On the first {} seeds, we have these average scores and win rates:\n\n", n_trials); let intro = format!("On the first {} seeds, we have these average scores and win rates:\n\n", n_trials);
let format_name = |x| format!(" {:7} ", x); let format_name = |x| format!(" {:7} ", x);
let format_players = |x| format!(" {}p ", x); let format_players = |x| format!(" {}p (stderr) ", x);
let format_percent = |x| format!(" {:05.2} % ", x); let format_percent = |x, stderr| format!(" {:05.2} % ({:.2} %) ", x, stderr);
let format_score = |x| format!(" {:07.4} ", x); let format_score = |x, stderr| format!(" {:07.4} ({:.4}) ", x, stderr);
let space = String::from(" "); let space = String::from(" ");
let dashes = String::from("---------"); let dashes = String::from("---------");
let dashes_long = String::from("------------------");
type TwoLines = (String, String); type TwoLines = (String, String);
fn make_twolines(player_nums: &Vec<u32>, head: TwoLines, make_block: &dyn Fn(u32) -> TwoLines) -> TwoLines { fn make_twolines(player_nums: &Vec<u32>, head: TwoLines, make_block: &dyn Fn(u32) -> TwoLines) -> TwoLines {
let mut blocks = player_nums.iter().cloned().map(make_block).collect::<Vec<_>>(); let mut blocks = player_nums.iter().cloned().map(make_block).collect::<Vec<_>>();
@ -185,11 +186,16 @@ fn get_results_table() -> String {
fn concat_twolines(body: Vec<TwoLines>) -> String { fn concat_twolines(body: Vec<TwoLines>) -> String {
body.into_iter().fold(String::default(), |output, (a, b)| (output + &a + "\n" + &b + "\n")) body.into_iter().fold(String::default(), |output, (a, b)| (output + &a + "\n" + &b + "\n"))
} }
let header = make_twolines(&player_nums, (space.clone(), dashes.clone()), &|n_players| (format_players(n_players), dashes.clone())); let header = make_twolines(&player_nums,
(space.clone(), dashes.clone()),
&|n_players| (format_players(n_players), dashes_long.clone()));
let mut body = strategies.iter().map(|strategy| { let mut body = strategies.iter().map(|strategy| {
make_twolines(&player_nums, (format_name(strategy), space.clone()), &|n_players| { make_twolines(&player_nums, (format_name(strategy), space.clone()), &|n_players| {
let simresult = sim_games(n_players, strategy, Some(seed), n_trials, n_threads, None); let simresult = sim_games(n_players, strategy, Some(seed), n_trials, n_threads, None);
(format_score(simresult.average_score()), format_percent(simresult.percent_perfect())) (
format_score(simresult.average_score(), simresult.score_stderr()),
format_percent(simresult.percent_perfect(), simresult.percent_perfect_stderr())
)
}) })
}).collect::<Vec<_>>(); }).collect::<Vec<_>>();
body.insert(0, header); body.insert(0, header);

View file

@ -97,6 +97,17 @@ impl Histogram {
pub fn average(&self) -> f32 { pub fn average(&self) -> f32 {
(self.sum as f32) / (self.total_count as f32) (self.sum as f32) / (self.total_count as f32)
} }
pub fn stdev_of_average(&self) -> f32 {
let average = self.average();
let mut var_sum = 0.0;
for (&val, &count) in self.hist.iter() {
var_sum += (val as f32 - average).powi(2) * count as f32;
}
// Divide by (self.total_count - 1) estimate the variance of the distribution,
// then divide by self.total_count estimate the variance of the sample average,
// then take the sqrt to get the stdev.
(var_sum / (((self.total_count - 1) * self.total_count) as f32)).sqrt()
}
pub fn merge(&mut self, other: Histogram) { pub fn merge(&mut self, other: Histogram) {
for (val, count) in other.hist.into_iter() { for (val, count) in other.hist.into_iter() {
self.insert_many(val, count); self.insert_many(val, count);
@ -196,10 +207,20 @@ impl SimResult {
self.scores.percentage_with(&PERFECT_SCORE) * 100.0 self.scores.percentage_with(&PERFECT_SCORE) * 100.0
} }
pub fn percent_perfect_stderr(&self) -> f32 {
let pp = self.percent_perfect() / 100.0;
let stdev = (pp*(1.0 - pp) / ((self.scores.total_count - 1) as f32)).sqrt();
stdev * 100.0
}
pub fn average_score(&self) -> f32 { pub fn average_score(&self) -> f32 {
self.scores.average() self.scores.average()
} }
pub fn score_stderr(&self) -> f32 {
self.scores.stdev_of_average()
}
pub fn average_lives(&self) -> f32 { pub fn average_lives(&self) -> f32 {
self.lives.average() self.lives.average()
} }